Patch [perl #111400] [:upper:] broken for above Latin1

This was an off-by-one error caused by my failing to realize that things
had to be done differently at the 255/256 boundary depending on whether
U+00FF matched or did not match the property.

Two properties were affected, [:upper:] and [:punct:].  The bug was that
all code points above the first one > 255 that legitimately matches the
property will match whether or not they should.  In the case of
[:upper:], this meant that effectively anything from 256..infinity
matched.  For [:punct:], it was anything above U+037D.

3 files changed, 51 insertions, 11 deletions

diff --git a/charclass_invlists.h b/charclass_invlists.h
index 152793a2e6..368410fd37 100644
--- a/charclass_invlists.h
+++ b/charclass_invlists.h
@@ -359,7 +359,7 @@ UV PosixPunct_invlist[] = {
 };
 
 UV L1PosixPunct_invlist[] = {
-	21,	/* Number of elements */
+	20,	/* Number of elements */
 	0,	/* Current iteration position */
 	1064334010, /* Version and data structure type */
 	1,	/* 0 if this is the first element of the list proper;
@@ -383,8 +383,7 @@ UV L1PosixPunct_invlist[] = {
 	187,
 	188,
 	191,
-	192,
-	894
+	192
 };
 
 UV PosixSpace_invlist[] = {
@@ -440,7 +439,7 @@ UV PosixUpper_invlist[] = {
 };
 
 UV L1PosixUpper_invlist[] = {
-	7,	/* Number of elements */
+	6,	/* Number of elements */
 	0,	/* Current iteration position */
 	1064334010, /* Version and data structure type */
 	1,	/* 0 if this is the first element of the list proper;
@@ -450,8 +449,7 @@ UV L1PosixUpper_invlist[] = {
 	192,
 	215,
 	216,
-	223,
-	256
+	223
 };
 
 UV PosixWord_invlist[] = {
diff --git a/regen/mk_invlists.pl b/regen/mk_invlists.pl
index 8102c29cd3..64e3d435c1 100644
--- a/regen/mk_invlists.pl
+++ b/regen/mk_invlists.pl
@@ -127,12 +127,17 @@ for my $prop (qw(
 
     # For the Latin1 properties, we change to use the eXtended version of the
     # base property, then go through the result and get rid of everything not
-    # in Latin1 (above 255).  Actually, we retain the element that crosses the
-    # 255/256 boundary.  For example, in the Word property, there is a range
-    # of code points that start at U+00F8 and goes through U+02C1.  Instead of
+    # in Latin1 (above 255).  Actually, we retain the element for the range
+    # that crosses the 255/256 boundary if it is one that matches the
+    # property.  For example, in the Word property, there is a range of code
+    # points that start at U+00F8 and goes through U+02C1.  Instead of
     # artifically cutting that off at 256 because 256 is the first code point
     # above Latin1, we let the range go to its natural ending.  That gives us
-    # extra information with no added space taken.
+    # extra information with no added space taken.  But if the range that
+    # crosses the boundary is one that doesn't match the property, we don't
+    # start a new range above 255, as that could be construed as going to
+    # infinity.  For example, the Upper property doesn't include the character
+    # at 255, but does include the one at 256.  We don't include the 256 one.
     my $lookup_prop = $prop;
     $lookup_prop =~ s/^L1Posix/XPosix/ or $lookup_prop =~ s/^L1//;
     my @invlist = prop_invlist($lookup_prop);
@@ -140,7 +145,22 @@ for my $prop (qw(
     if ($lookup_prop ne $prop) {
         for my $i (0 .. @invlist - 1 - 1) {
             if ($invlist[$i] > 255) {
-                splice @invlist, $i+1;
+
+                # In an inversion list, even-numbered elements give the code
+                # points that begin ranges that match the property;
+                # odd-numbered give ones that begin ranges that don't match.
+                # If $i is odd, we are at the first code point above 255 that
+                # doesn't match, which means the range it is ending does
+                # match, and crosses the 255/256 boundary.  We want to include
+                # this ending point, so increment $i, so the splice below
+                # includes it.  Conversely, if $i is even, it is the first
+                # code point above 255 that matches, which means there was no
+                # matching range that crossed the boundary, and we don't want
+                # to include this code point, so splice before it.
+                $i++ if $i % 2 != 0;
+
+                # Remove everything past this.
+                splice @invlist, $i;
                 last;
             }
         }
diff --git a/t/re/re_tests b/t/re/re_tests
index e7680c9c23..4d78a6aed8 100644
--- a/t/re/re_tests
+++ b/t/re/re_tests
@@ -1575,4 +1575,26 @@ abc\N{def	-	c	-	\\N{NAME} must be resolved by the lexer
 # [perl #110648]
 [^\p{Alphabetic}]	\x{100}	n	-	-
 
+# [perl #111400].  Tests the first Y/N boundary above 255 for each of these.
+/[[:alnum:]]/	\x{2c1}	y	-	-
+/[[:alnum:]]/	\x{2c2}	n	-	-
+/[[:alpha:]]/	\x{2c1}	y	-	-
+/[[:alpha:]]/	\x{2c2}	n	-	-
+/[[:graph:]]/	\x{377}	y	-	-
+/[[:graph:]]/	\x{378}	n	-	-
+/[[:lower:]]/	\x{100}	n	-	-
+/[[:lower:]]/	\x{101}	y	-	-
+/[[:lower:]]/	\x{102}	n	-	-
+/[[:print:]]/	\x{377}	y	-	-
+/[[:print:]]/	\x{378}	n	-	-
+/[[:punct:]]/	\x{37D}	n	-	-
+/[[:punct:]]/	\x{37E}	y	-	-
+/[[:punct:]]/	\x{388}	n	-	-
+/[[:upper:]]/	\x{100}	y	-	-
+/[[:upper:]]/	\x{101}	n	-	-
+/[[:word:]]/	\x{2c1}	y	-	-
+/[[:word:]]/	\x{2c2}	n	-	-
+
+
+
 # vim: softtabstop=0 noexpandtab